Convert an input string to a regex without $eval? (or, more generally, non-literal regex support)

[mcskatkat]

So JSONata is now the official AWS StepFunctions query language and that's great. It's also how I got to know and appreciate it.

To be usable in such a context it has (justifiably) lost $eval() (in lieu of which AWS provides $parse(), to deserialize strings containing inert JSON). I am not mourning $eval as I am generally against dynamic parsing / recursive interpreter invocations, for inevitably introducing security / reliability problems in production-grade environments.

However, barring $eval and with JSONata's design choice that a regex/pattern is a function, it seems there is no way to convert an inert JSON string such as "^s3://my-bucket/" to a regular expression that can be matched against a runtime string.
(I am ignoring, of course, the theoretical possibility of using JSONata's Turing completeness to implement a regular-expression matching engine 😕).

Equivalently, without $eval (arguably, even with it), JSONata's regex support seems to be intended for literal expressions only, and is mostly incapable of programmatic manipulation of regular expressions (e.g. combine $prefix and $suffix to create "^" & $prefix & ".*" & $suffix & "$"), which is possible and useful in many contexts.

Here's one possible, fairly straightforward solution ( = enhancement request):

• $prefix_re := $regex("s3://my-bucket/")
  same as /s3:\/\/my-bucket\//, no need to escape the all-too-common slash

• $suffix_re := $regex($regex_escape(user_provided_suffix))
$regex_escape() is pretty essential too

• $prefix_re_source := $prefix_re()
  would be nice if this returns "^s3://my-bucket/" rather than ** no match **

• $full_uri_re := $regex("^" & $prefix_re() & ".*" & $suffix_re() & "$")
  regex compounding

• user_provided_uris[$contains($full_uri_re)]
  compound, non-literal regex used like any other regex

Would be glad to hear people's opinions on this.

[markmelville]

The prefix and suffix examples don't need the full power of regex and could utilize a startsWith or endsWith shim. Where you do really need the full power of a regex, I would normally say use jsonata's extensibility features to invoke the actual RegExp constructor, but I don't believe the extend function is accessible in a Step Functions context.